raphiaura/KytyPS5
1
0
Fork 0
mirror of https://github.com/KytyPS5/KytyPS5.git synced 2026-10-20 10:56:58 +02:00
Code Issues Packages Projects Releases Wiki Activity

shader: preserve RDNA2 DLC coherence for flat address loads

Browse Source
This commit is contained in:
Claxten
2026-10-06 03:28:30 +04:00
committed by nmzik
parent cdb64bfd4f
commit 9caec9f2c8
5 changed files with 75 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/graphics/shader/recompiler/frontend/decode/MemoryOps.cpp
+2 -2
View File
@@ -343,7 +343,6 @@ void DecodeFlat(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
const uint32_t word0 = code[word_index];
const uint32_t word1 = code[word_index + 1u];
const uint32_t offset = word0 & 0xfffu;
const uint32_t dlc = (word0 >> 12u) & 1u;
const uint32_t lds = (word0 >> 13u) & 1u;
const uint32_t seg = (word0 >> 14u) & 0x3u;
const uint32_t opcode = (word0 >> 18u) & 0x7fu;
@@ -355,6 +354,7 @@ void DecodeFlat(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
inst.pc = pc;
inst.word_count = 2;
inst.offset = seg == 0u ? (offset & 0x7ffu) : SignExtendU32(offset, 12u);
inst.dlc = ((word0 >> 12u) & 1u) != 0;
inst.glc = ((word0 >> 16u) & 1u) != 0;
inst.slc = ((word0 >> 17u) & 1u) != 0;
inst.family = Family::FLAT;
@@ -364,7 +364,7 @@ void DecodeFlat(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
ApplyMemoryInfo(inst, info);
SetRawWords(inst, code, word_index, 2);
if (dlc != 0 || lds != 0 || inst.glc || inst.slc || seg == 3u) {
if (lds != 0 || inst.glc || inst.slc || seg == 3u) {
SetUnsupported(inst, Family::FLAT, opcode, "FLAT modifiers or segment are not implemented");
return;
}
src/graphics/shader/recompiler/frontend/translate/Memory.cpp
+3 -2
View File
@@ -129,8 +129,9 @@ IR::MemoryInfo MemoryInfoFromDecoded(const Decoder::Instruction& decoded) {
// Vector loads use GLC/DLC to bypass L0/GL1; atomics use GLC only to return data.
const bool buffer_atomic = decoded.opcode >= Decoder::Opcode::BUFFER_ATOMIC_SWAP &&
decoded.opcode <= Decoder::Opcode::BUFFER_ATOMIC_FMAX;
memory.coherent = memory.kind == ResourceKind::Buffer && !buffer_atomic &&
(decoded.glc || decoded.dlc);
memory.coherent = (memory.kind == ResourceKind::Buffer || memory.kind == ResourceKind::Flat ||
memory.kind == ResourceKind::Global) &&
!buffer_atomic && (decoded.glc || decoded.dlc);
memory.resource = ResourceIndexFromOperand(decoded.src1);
memory.sampler = ResourceIndexFromOperand(decoded.src2);
if (memory.kind == ResourceKind::ScalarBuffer) {